How this was built

hiren8 is a small, intentional stack for talking about Nate’s background: a Next.js conversation UI streams answers from a FastAPI service that grounds Gemini Flash in a curated markdown pack — not the open web.

Your browser only talks to this site. A server-side Route Handler proxies the Server-Sent Events stream to the API. In production the API stays private on Cloud Run (IAM); the web service attaches an identity token. Locally it is two processes and plain HTTP.

Orchestration is LangChain LCEL. There is no citation chrome — the chat stage is the product. Caps, evals, and deploy live in the repo for anyone who wants the operator view.